rasa-pro 3.14.1__py3-none-any.whl → 3.15.0a3__py3-none-any.whl

rasa/builder/service.py

@@ -5,6 +5,7 @@ import time
 from http import HTTPStatus
 from typing import Any, Optional
 
+import langfuse
 import structlog
 from sanic import Blueprint, HTTPResponse, response
 from sanic.request import Request
@@ -41,7 +42,6 @@ from rasa.builder.copilot.signing import (
     create_signature_envelope_for_text,
     verify_signature,
 )
-from rasa.builder.copilot.telemetry import CopilotTelemetry
 from rasa.builder.download import create_bot_project_archive
 from rasa.builder.guardrails.constants import (
     BLOCK_SCOPE_PROJECT,
@@ -51,6 +51,7 @@ from rasa.builder.guardrails.constants import (
 from rasa.builder.guardrails.store import guardrails_store
 from rasa.builder.job_manager import job_manager
 from rasa.builder.jobs import (
+    run_backup_to_bot_job,
     run_prompt_to_bot_job,
     run_replace_all_files_job,
     run_template_to_bot_job,
@@ -65,15 +66,19 @@ from rasa.builder.models import (
     ApiErrorResponse,
     AssistantInfo,
     BotData,
+    BotFiles,
     JobCreateResponse,
     JobStatus,
     JobStatusEvent,
     PromptRequest,
+    RestoreFromBackupRequest,
     ServerSentEvent,
     TemplateRequest,
 )
 from rasa.builder.project_generator import ProjectGenerator
 from rasa.builder.shared.tracker_context import TrackerContext
+from rasa.builder.telemetry.copilot_langfuse_telemetry import CopilotLangfuseTelemetry
+from rasa.builder.telemetry.copilot_segment_telemetry import CopilotSegmentTelemetry
 from rasa.core.agent import Agent
 from rasa.core.channels.studio_chat import StudioChatInput
 from rasa.core.exceptions import AgentNotReady
@@ -463,6 +468,95 @@ async def handle_template_to_bot(request: Request) -> HTTPResponse:
         )
 
 
+@bp.route("/backup-to-bot", methods=["POST"])
+@openapi.summary("Generate bot from backup archive")
+@openapi.description(
+    "Creates a complete conversational AI bot from a backup tar.gz archive. "
+    "Returns immediately with a job ID. Connect to `/job-events/<job_id>` to "
+    "receive server-sent events (SSE) for real-time progress tracking "
+    "throughout the bot restoration process.\n\n"
+    "**SSE Event Flow** (via `/job-events/<job_id>`):\n"
+    "1. `received` - Request received by server\n"
+    "2. `generating` - Extracting and restoring bot from backup\n"
+    "3. `generation_success` - Backup restoration completed successfully\n"
+    "4. `training` - Training the bot model (if no existing model found)\n"
+    "5. `train_success` - Model training completed (if training was needed)\n"
+    "6. `done` - Bot restoration completed\n\n"
+    "**Error Events:**\n"
+    "- `generation_error` - Failed to restore bot from backup\n"
+    "- `train_error` - Backup restored but training failed\n"
+    "- `validation_error` - Restored bot configuration is invalid\n"
+    "- `error` - Unexpected error occurred\n\n"
+    "**Usage:**\n"
+    "1. Send POST request with Content-Type: application/json\n"
+    "2. The response will be a JSON object `{job_id: ...}`\n"
+    "3. Connect to `/job-events/<job_id>` for a server-sent event stream of progress."
+)
+@openapi.tag("bot-generation")
+@openapi.body(
+    {"application/json": model_to_schema(RestoreFromBackupRequest)},
+    description="Backup request with presigned URL to tar.gz archive.",
+    required=True,
+    example={"presigned_url": "https://s3.amazonaws.com/bucket/path?signature=..."},
+)
+@openapi.response(
+    200,
+    {"application/json": model_to_schema(JobCreateResponse)},
+    description="Job created. Poll or subscribe to /job-events/<job_id> for progress.",
+)
+@openapi.response(
+    400,
+    {"application/json": model_to_schema(ApiErrorResponse)},
+    description="Validation error in request payload or invalid presigned URL",
+)
+@openapi.response(
+    500,
+    {"application/json": model_to_schema(ApiErrorResponse)},
+    description="Internal server error",
+)
+@openapi.parameter(
+    HEADER_USER_ID,
+    description=(
+        "Optional user id to associate requests (e.g., for telemetry/guardrails)."
+    ),
+    _in="header",
+    required=False,
+    schema=str,
+)
+async def handle_backup_to_bot(request: Request) -> HTTPResponse:
+    """Handle backup-to-bot restoration requests."""
+    try:
+        payload = RestoreFromBackupRequest(**request.json)
+    except Exception as exc:
+        return response.json(
+            ApiErrorResponse(
+                error="Invalid request", details={"error": str(exc)}
+            ).model_dump(),
+            status=400,
+        )
+
+    try:
+        # Allocate job and schedule background task
+        job = job_manager.create_job()
+        request.app.add_task(
+            run_backup_to_bot_job(request.app, job, payload.presigned_url)
+        )
+        return response.json(JobCreateResponse(job_id=job.id).model_dump(), status=200)
+    except Exception as exc:
+        capture_exception_with_context(
+            exc,
+            "bot_builder_service.backup_to_bot.unexpected_error",
+            tags={"endpoint": "/api/backup-to-bot"},
+        )
+        return response.json(
+            ApiErrorResponse(
+                error="Failed to create backup-to-bot job",
+                details={"error": str(exc)},
+            ).model_dump(),
+            status=HTTPStatus.INTERNAL_SERVER_ERROR,
+        )
+
+
 @bp.route("/files", methods=["GET"])
 @openapi.summary("Get bot files")
 @openapi.description(
@@ -798,16 +892,14 @@ async def get_bot_info(request: Request) -> HTTPResponse:
 @openapi.summary("Download bot project as tar.gz")
 @openapi.description(
     "Downloads the current bot project files as a compressed tar.gz archive. "
-    "Includes all configuration files and a .env file with RASA_PRO_LICENSE. "
-    "Requires valid JWT token in Authorization header."
+    "Includes all configuration files and a .env file with RASA_PRO_LICENSE."
 )
 @openapi.tag("bot-files")
 @openapi.parameter(
-    "Authorization",
-    description=("Bearer token for authentication. Always required for this endpoint."),
-    _in="header",
-    required=True,
-    schema=str,
+    "exclude_models_directory",
+    bool,
+    location="query",
+    description="Whether to exclude the models directory",
 )
 @openapi.parameter(
     HEADER_USER_ID,
@@ -830,32 +922,25 @@ async def get_bot_info(request: Request) -> HTTPResponse:
     {"application/gzip": bytes},
     description="Bot project downloaded successfully as tar.gz",
 )
-@openapi.response(
-    401,
-    {"application/json": model_to_schema(ApiErrorResponse)},
-    description=(
-        "Authentication failed - Authorization header missing or invalid. "
-        "Authentication is always required for this endpoint."
-    ),
-)
 @openapi.response(
     500,
     {"application/json": model_to_schema(ApiErrorResponse)},
     description="Internal server error",
 )
-@protected(always_required=True)
 async def download_bot_project(request: Request) -> HTTPResponse:
     """Download bot project as tar.gz archive."""
     try:
-        # Token verification is enforced by the
-        # protected(always_required=True) decorator.
+        # Get query parameters
+        exclude_models_directory = (
+            request.args.get("exclude_models_directory", "true").lower() == "true"
+        )
+        project_name = request.args.get("project_name", "bot-project")
 
         # Get bot files
         project_generator = get_project_generator(request)
-        bot_files = project_generator.get_bot_files()
-
-        # Get project name from query parameters, default to "bot-project"
-        project_name = request.args.get("project_name", "bot-project")
+        bot_files = project_generator.get_bot_files(
+            exclude_models_directory=exclude_models_directory
+        )
 
         # Create tar.gz archive
         tar_data = create_bot_project_archive(bot_files, project_name)
@@ -1020,6 +1105,9 @@ async def download_bot_project(request: Request) -> HTTPResponse:
     schema=str,
 )
 @protected()
+# Disable automatic input/output capture for langfuse tracing
+# This allows manual control over what data is sent to langfuse
+@langfuse.observe(capture_input=False, capture_output=False)
 async def copilot(request: Request) -> None:
     """Handle copilot requests with streaming markdown responses."""
     sse = await request.respond(content_type="text/event-stream")
@@ -1046,9 +1134,12 @@ async def copilot(request: Request) -> None:
             )
             return
 
-        telemetry = CopilotTelemetry(project_id=HELLO_RASA_PROJECT_ID, user_id=user_id)
+        telemetry = CopilotSegmentTelemetry(
+            project_id=HELLO_RASA_PROJECT_ID, user_id=user_id
+        )
         structlogger.debug("builder.copilot.telemetry.request.init")
 
+        # TODO: This can be removed once Langfuse is completed.
         if req.last_message and req.last_message.role == ROLE_USER:
             structlogger.debug("builder.copilot.telemetry.request.user_turn")
             # Offload telemetry logging to a background task
@@ -1088,26 +1179,9 @@ async def copilot(request: Request) -> None:
             return
 
         # 4. Get the necessary context for the copilot
-        tracker = await current_tracker_from_input_channel(request.app, req.session_id)
-        tracker_context = TrackerContext.from_tracker(
-            tracker, max_turns=COPILOT_ASSISTANT_TRACKER_MAX_TURNS
-        )
-        if (
-            tracker_context is not None
-            and llm_service.guardrails_policy_checker is not None
-        ):
-            tracker_context = await llm_service.guardrails_policy_checker.check_assistant_chat_for_policy_violations(  # noqa: E501
-                tracker_context=tracker_context,
-                hello_rasa_user_id=user_id,
-                hello_rasa_project_id=HELLO_RASA_PROJECT_ID,
-                lakera_project_id=LAKERA_ASSISTANT_HISTORY_GUARDRAIL_PROJECT_ID,
-            )
-
-        # Copilot doesn't need to know about the docs and any file that is not a core
-        # assistant file
-        relevant_assistant_files = project_generator.get_bot_files(
-            exclude_docs_directory=True,
-            allowed_file_extensions=["yaml", "yml", "py", "jinja", "jinja2"],
+        tracker_context = await get_tracker_context_for_copilot(request, req, user_id)
+        relevant_assistant_files = get_relevant_assistant_files_for_copilot(
+            project_generator,
         )
         context = CopilotContext(
             tracker_context=tracker_context,
@@ -1162,7 +1236,7 @@ async def copilot(request: Request) -> None:
         async for token in intercepted_stream:
             await sse.send(token.to_sse_event().format())
 
-        # 8. Offload telemetry logging to a background task
+        # 8a. Offload metabase telemetry logging to a background task
         request.app.add_task(
             asyncio.to_thread(
                 telemetry.log_copilot_from_handler,
@@ -1177,9 +1251,27 @@ async def copilot(request: Request) -> None:
                     else None
                 ),
                 tracker_event_attachments=generation_context.tracker_event_attachments,
-                **copilot_client.usage_statistics.model_dump(),
+                model=copilot_client.usage_statistics.model or "N/A",
+                prompt_tokens=copilot_client.usage_statistics.prompt_tokens or 0,
+                cached_prompt_tokens=(
+                    copilot_client.usage_statistics.cached_prompt_tokens or 0
+                ),
+                completion_tokens=(
+                    copilot_client.usage_statistics.completion_tokens or 0
+                ),
+                total_tokens=copilot_client.usage_statistics.total_tokens or 0,
             )
         )
+        # 8b. Setup output trace attributes for Langfuse
+        CopilotLangfuseTelemetry.setup_copilot_endpoint_call_trace_attributes(
+            hello_rasa_project_id=HELLO_RASA_PROJECT_ID or "N/A",
+            chat_id=req.session_id or "N/A",
+            user_id=user_id,
+            request=req,
+            handler=copilot_response_handler,
+            relevant_documents=generation_context.relevant_documents,
+            copilot_context=context,
+        )
 
         # 9. Once the stream is over, extract and send references
         #    if any documents were used
@@ -1365,3 +1457,70 @@ async def _handle_guardrail_violation_and_maybe_block(
 
     await sse.send(message.to_sse_event().format())
     return message
+
+
+@langfuse.observe(capture_input=False, capture_output=False)
+async def get_tracker_context_for_copilot(
+    request: Request,
+    req: CopilotRequest,
+    user_id: str,
+) -> Optional[TrackerContext]:
+    """Check the assistant chat for guardrail policy violations.
+
+    Args:
+        request: The request object.
+        req: The CopilotRequest object.
+        user_id: The user ID.
+
+    Returns:
+        The tracker context if the tracker is available.
+    """
+    tracker = await current_tracker_from_input_channel(request.app, req.session_id)
+    tracker_context = TrackerContext.from_tracker(
+        tracker, max_turns=COPILOT_ASSISTANT_TRACKER_MAX_TURNS
+    )
+    if (
+        tracker_context is not None
+        and llm_service.guardrails_policy_checker is not None
+    ):
+        tracker_context = await llm_service.guardrails_policy_checker.check_assistant_chat_for_policy_violations(  # noqa: E501
+            tracker_context=tracker_context,
+            hello_rasa_user_id=user_id,
+            hello_rasa_project_id=HELLO_RASA_PROJECT_ID,
+            lakera_project_id=LAKERA_ASSISTANT_HISTORY_GUARDRAIL_PROJECT_ID,
+        )
+
+    # Track the retrieved tracker context
+    CopilotLangfuseTelemetry.trace_copilot_tracker_context(
+        tracker_context=tracker_context,
+        max_conversation_turns=COPILOT_ASSISTANT_TRACKER_MAX_TURNS,
+        session_id=req.session_id,
+    )
+
+    return tracker_context
+
+
+@langfuse.observe(capture_input=False, capture_output=False)
+def get_relevant_assistant_files_for_copilot(
+    project_generator: ProjectGenerator,
+) -> BotFiles:
+    """Get the relevant assistant files for the copilot.
+
+    Args:
+        project_generator: The project generator.
+
+    Returns:
+        The relevant assistant files.
+    """
+    # Copilot doesn't need to know about the docs and any file that is not a core
+    # assistant file
+    files = project_generator.get_bot_files(
+        exclude_docs_directory=True,
+        allowed_file_extensions=["yaml", "yml", "py", "jinja", "jinja2"],
+    )
+
+    # Track the retrieved assistant files
+    CopilotLangfuseTelemetry.trace_copilot_relevant_assistant_files(
+        relevant_assistant_files=files,
+    )
+    return files